Seismic quality factor estimation using frequency-dependent linear fitting

The estimation of the quality factor (Q factor) is very important for seismic data attenuation compensation and hydrocarbon indication. If both random noise and seismic signal co-exist, the logarithmic spectrum ratio (LSR) method, known as a popular method for Q estimation, is usually unstable and brings large errors. In order to improve the accuracy and stability of LSR method, in this paper, we propose an improved method for Q estimation by introducing frequency-dependent linear fitting (FDLF) in the spectral ratio method. We first use nonstationary linear fitting to estimate many Q factors, and then utilize the median value of these estimated Q values as the final estimation of Q factor. Using median value can effectively overcome the difficulties in effective frequency band selection in LSR method. The estimated cumulative Q can be transformed into interval layer Q. The proposed method can help to obtain more precise and stable result especially when the seismic data contains noise. We use synthetic and field zero-offset vertical seismic profile data examples to illustrate our method.